Let a Good Chatbot Answer That Question

There’s a classic story that brings together two of the most notorious American political figures of the 20th century, Richard Nixon and Joe McCarthy, and the syndicated muckraker who did as much as anyone to expose their transgressions, Washington Merry-Go-Round columnist Drew Pearson. The story appears with slight variations in four different books by my count, most recently in Donald A. Ritchie’s 2021 Pearson biography The Columnist.

The incident occurred during a December 1950 soirée at Washington, D.C.’s posh Sulgrave Club, when McCarthy and Pearson spent much of the evening trading barbs and nearly coming to blows before wives and friends intervened. They collided one last time in the Sulgrave cloakroom, where a roaring drunk McCarthy pounded the columnist’s head and kneed him twice in the groin. Nixon, a new senator, arrived just in time to see McCarthy deliver another “head-snapping” slap to Pearson. Nixon stepped between the combatants and said, “Let a good Quaker break up this fight.” Nixon then walked a stumbling McCarthy out of the club and spent the next hour patiently searching the surrounding streets for McCarthy’s car, before helping his friend into the driver’s seat and sending him merrily on his way.

The Columnist tells the tamest version of the story, and although it doesn’t even mention Nixon’s peacemaker line, the book sheds light on its apparent hale-fellow bonhomie that no other rendering does. Nixon and Pearson were both raised Quaker, but resentful targets of Pearson’s many career-crushing exposés liked to claim that the columnist was a fake Quaker who had opportunistically “thee’d and thou’d” his way out of serving in World War I. Knowing Nixon as a champion grudge-holder (and that Pearson was a charter member of Nixon’s infamous enemies list), it’s hard to miss the likelihood that Nixon was mocking Pearson with the “good Quaker” line even as he was ostensibly saving him from McCarthy’s below-the-belt assault.

What’s fascinating about the multiple accounts is that they don’t so much contradict each other as reveal different dimensions of the incident and about the three consistently captivating (if not especially admirable) historical figures involved. Although I couldn’t quite figure out an effective way to ask it the question, with all of the recent chatter about OpenAI’s ChatGPT and the conviction of several writers I know that the AI conversational language bot will soon put them out to pasture, I wondered if ChatGPT could ever deliver the sort of context-enriched nuance that human historians provide through different research methods, analysis, perspectives, and value assessments of which details of a story merit inclusion and which ones don’t.

I’ve seen a couple of pretty interesting Chat­GPT experiments pertaining to Streaming Media’s purview—queries on the differences between HLS and DASH or the relative merits of AV1 and HEVC and why or why not to deploy different codecs. The results appear accurate, although just a nuance or two off on some important points. So far, two unsolicited AI-generated “What Is Streaming?”-type articles have reached my inbox, passed off as a human writer’s work. Neither was egregiously bad, but both were unmistakably missing the deep an­aly­sis and broad perspective that Streaming Media contributing editors supply in the definitive state-of-the-industry reports that populate this year’s 20th annual Sourcebook.

If the recoiling-in-fear “Is AI coming for our jobs?” question is premature, it’s also far from the most interesting topic these experiments raise. Rather, it’s how can we get better answers by asking better questions? Just as I write this in late January, the Wharton School of Business has announced that the bot scored an unassisted B on an M.B.A. final exam. What ChatGPT can do today is remarkable. And it’s a still-iterating product. The version OpenAI released to the world is only ChatGPT 3.5 (and the version that took the Wharton test was “only” 3.0).

What’s more, the mastery of nuance the bot currently lacks is by definition a game of inches. And it’s not as if humans communicate every subtle shade of meaning every time. Given that it took the combined efforts of no fewer than four historians to reveal the precise meaning of Nixon’s seemingly innocuous “good Quaker” jab at Drew Pearson, who knows what the cumulative virtuosity of four or more iterations of ChatGPT will deliver?

What Else Did We Get Wrong?

Based on what I'm hearing from a wide array of streaming producers, the heightened demand for streaming live events that we expected to be a natural outcome of its COVID-era ascendancy is either evaporating or simply hasn't materialized.

07 Jun 2023